[Mesa-dev] [PATCH v5 0/3] A couple of fixes for Tex(Sub)Image error checks

[Jason Ekstrand]

On Fri, Sep 25, 2015 at 9:06 AM, Crisan, Alex <alex.crisan at intel.com> wrote:
> Sure Mark, I will work with Kaveh on the next steps. Is it safe to assume these tests will be buggy on any platforms?

Yes.  The test is testing for GL API errors and its behavior should be
completely independent of platform.
--Jason

> -----Original Message-----
> From: Janes, Mark A
> Sent: Friday, September 25, 2015 7:41 AM
> To: Eduardo Lima Mitev; mesa-dev at lists.freedesktop.org
> Cc: Eduardo Lima Mitev; jason at jlekstrand.net; Nasri, Kaveh; Mason, Michael W; Crisan, Alex
> Subject: Re: [PATCH v5 0/3] A couple of fixes for Tex(Sub)Image error checks
>
> This patch series fixes a few dEQP tests on braswell.  The rest of the platform are unaffected, according to my tests.
>
> deqp-gles2.functional.negative_api.texture.teximage2d_invalid_type
> deqp-gles2.functional.negative_api.texture.teximage2d_invalid_internalformat
> deqp-gles3.functional.negative_api.texture.teximage2d
>
> Since the dEQP regressions listed in the bug have been found to be due to a poorly written test, someone will need to take the action to resolve this discrepancy with Google.  I don't expect AOSP to be the project which can resolve this test bug.
>
> Jason, Alex, and Kaveh: can you figure out how to address this issue?
>
> -Mark
>
> Eduardo Lima Mitev <elima at igalia.com> writes:
>
>> This is a new version of the series that attempt to fix the regression reported at:
>>
>> https://bugs.freedesktop.org/show_bug.cgi?id=91582
>>
>> The review by Jason helped me uncover the fact that the following 3 dEQP tests are buggy:
>>
>> dEQP-GLES2.functional.negative_api.texture.texsubimage2d_neg_offset
>> dEQP-GLES2.functional.negative_api.texture.texsubimage2d_offset_allowe
>> d dEQP-GLES2.functional.negative_api.texture.texsubimage2d_neg_wdt_hgt
>>
>> So the patch split I did in the previous version of the series (v4) was actually not necessary. It was just a work-around to the failure of these tests, which this series uncovered.
>>
>> Now in this new version, I dropped the splitted patch and filed a bug against dEQP (together with a reference patch) to fix the above tests, which will start to fail once/if we merge this series.
>>
>> "[dEQP] Buggy negative API tests that check dimensions args of
>> glTexSubImage2D"
>> <https://code.google.com/p/android/issues/detail?id=187348&thanks=1873
>> 48&ts=1443083425>
>>
>> I filed it against the AOSP project, where dEQP package is (under external/deqp). Lets see if that was correct.
>>
>> Mark, in the mean time we can probably apply the patch I attached to the bug report, otherwise the regression originally reported won't go away. What do you think?
>>
>> Notice that first two patches has R-b from Jason already. Only the patch 3/3 is pending review.
>>
>> The question that remains is whether I should cc Mesa 10.6 stable too, apart from 11.0.
>>
>> cheers,
>>
>> Eduardo Lima Mitev (3):
>>   mesa: Fix order of format+type and internal format checks for
>>     glTexImageXD ops
>>   mesa: Move _mesa_base_tex_format() from teximage to glformats files
>>   mesa: Use the effective internal format instead for validation
>>
>>  src/mesa/main/glformats.c | 656 ++++++++++++++++++++++++++++++++++++++++++++++
>>  src/mesa/main/glformats.h |   2 +
>>  src/mesa/main/teximage.c  | 415 ++---------------------------
>>  src/mesa/main/teximage.h  |   4 -
>>  4 files changed, 683 insertions(+), 394 deletions(-)
>>
>> --
>> 2.4.6